Panel data between estimator

[Alfonso Sánchez-Peñalver]

Hello everyone,

I was preparing a little example to show my students how the between estimator is simply an estimator on the averages of the different variables at the panel level, and curiously enough I found that Stata provides close but different in the example I was preparing. Here is the code:

clear all
set more off

webuse nlswork, clear
global x "age race collgrad grade hours"
keep idcode year ln_wage $x

* xtreg be
xtset idcode year
xtreg ln_wage $x, be

* regress estimation
collapse ln_wage $x, by(idcode)
regress ln_wage $x

My question is why is there this difference in the estimation of the coefficients? Is this because the panel is unbalanced and when using xtreg, be Stata weights the observations within the panels differently to estimate the panel average? If so, how does it calculate these weights?

Thanks,

Alfonso.

[Scott Merryman]

Use the casewise deletion option

collapse ln_wage $x , by(idcode) cw

[Alfonso Sánchez-Peñalver]

Thank you Scott, so yes it was because of missing observations, and thus unbalanced panels. Again thanks.